readygpu

GLM 4.7 Flash

31.2B total, 5.2B active · 198k context · mit · released 2026-01-19

codingagentic

What to run it on in the US

Fully in memory at 128k context, priced at retailers today.

Something we price holds it fully in memory, but none reaches 15 tok/s, our bar for recommending it. The table below has the speeds.

All hardware that runs it

Everything we price, best value first. Two cards split the model by layers.

HardwareQuantFitsDecodePricetok/s per 1,000
NVIDIA GeForce RTX 5090Q6_Kfits60 tok/s––
Mac Studio M4 Max 128GBQ8_0fits17 tok/s––
Mac mini M4 Pro 64GBQ8_0fits9 tok/s––
Framework Desktop Ryzen AI Max+ 395 128GBQ8_0fits7 tok/s––
Mac Studio M3 Ultra 512GBQ8_0fits26 tok/s––
NVIDIA DGX Spark 128GBQ8_0fits8 tok/s––
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBQ8_0fits7 tok/s––
Mac Studio M3 Ultra 256GBQ8_0fits26 tok/s––
Mac mini M4 Pro 48GBQ6_Kfits10 tok/s––
Mac Studio M4 Max 64GBQ8_0fits17 tok/s––
Mac Studio M3 Ultra 96GBQ8_0fits26 tok/s––
NVIDIA RTX PRO 6000 BlackwellQ8_0fits54 tok/s––
AMD Radeon AI PRO R9700Q6_Kfits19 tok/s––
ASUS Ascent GX10 128GBQ8_0fits8 tok/s––
Framework Desktop Ryzen AI Max+ 395 64GBQ8_0fits7 tok/s––
NVIDIA GeForce RTX 3090Q4_K_Moffload20 tok/s––
NVIDIA GeForce RTX 4090Q4_K_Moffload21 tok/s––
NVIDIA GeForce RTX 5070 TiQ4_K_Moffload7 tok/s––
NVIDIA GeForce RTX 5060 Ti 16GBQ4_K_Moffload6 tok/s––
AMD Radeon RX 7900 XTXQ4_K_Moffload18 tok/s––
NVIDIA GeForce RTX 3060 12GBQ4_K_Moffload5 tok/s––
AMD Radeon RX 9070 XTQ4_K_Moffload6 tok/s––
NVIDIA GeForce RTX 4060 Ti 16GBQ4_K_Moffload5 tok/s––
NVIDIA GeForce RTX 5080Q4_K_Moffload7 tok/s––
NVIDIA GeForce RTX 4070 Ti SuperQ4_K_Moffload7 tok/s––
NVIDIA GeForce RTX 5070Q4_K_Moffload5 tok/s––
AMD Radeon RX 9060 XT 16GBQ4_K_Moffload5 tok/s––
NVIDIA GeForce RTX 4070 SuperQ4_K_Moffload5 tok/s––
NVIDIA GeForce RTX 4080 SuperQ4_K_Moffload7 tok/s––
AMD Radeon RX 7800 XTQ4_K_Moffload6 tok/s––
AMD Radeon RX 7900 XTQ4_K_Moffload9 tok/s––
Intel Arc B580Q4_K_Moffload4 tok/s––
NVIDIA GeForce RTX 3090 TiQ4_K_Moffload21 tok/s––
AMD Radeon RX 7600 XTQ4_K_Moffload5 tok/s––
NVIDIA GeForce RTX 3080 10GBQ4_K_Moffload5 tok/s––
Intel Arc Pro B60Q4_K_Moffload10 tok/s––
NVIDIA Tesla P40Q4_K_Moffload10 tok/s––
Intel Arc A770 16GBQ4_K_Moffload6 tok/s––

2 other cards and machines we price cannot hold it at this context.

Memory needed

Weights, KV cache and runtime buffers at each context length (KV cache at fp16).

QuantFile@8k@32k@64k@128k
Q4_K_Mgguf17.1 GB18.2 GB19.6 GB21.4 GB25.1 GB
mlx-4bitmlx15.7 GB16.9 GB18.2 GB20.1 GB23.8 GB
Q5_K_Mgguf19.9 GB21.1 GB22.5 GB24.3 GB28.0 GB
Q6_Kgguf23.0 GB24.2 GB25.5 GB27.4 GB31.1 GB
mlx-8bitmlx29.6 GB30.8 GB32.2 GB34.0 GB37.7 GB
Q8_0gguf29.7 GB30.8 GB32.2 GB34.0 GB37.7 GB

Run it

Copy a command for the runtime you use.

  • Q4_K_M
    • Ollamaollama run hf.co/unsloth/GLM-4.7-Flash-GGUF:Q4_K_M
    • LM Studiolms get unsloth/GLM-4.7-Flash-GGUF@q4_k_m
    • llama.cppllama-server -hf unsloth/GLM-4.7-Flash-GGUF:Q4_K_M
  • mlx-4bit
    • LM Studiolms get mlx-community/GLM-4.7-Flash-4bit
    • MLXmlx_lm.server --model mlx-community/GLM-4.7-Flash-4bit
  • Q5_K_M
    • Ollamaollama run hf.co/unsloth/GLM-4.7-Flash-GGUF:Q5_K_M
    • LM Studiolms get unsloth/GLM-4.7-Flash-GGUF@q5_k_m
    • llama.cppllama-server -hf unsloth/GLM-4.7-Flash-GGUF:Q5_K_M
  • Q6_K
    • Ollamaollama run hf.co/unsloth/GLM-4.7-Flash-GGUF:Q6_K
    • LM Studiolms get unsloth/GLM-4.7-Flash-GGUF@q6_k
    • llama.cppllama-server -hf unsloth/GLM-4.7-Flash-GGUF:Q6_K
  • mlx-8bit
    • LM Studiolms get mlx-community/GLM-4.7-Flash-8bit
    • MLXmlx_lm.server --model mlx-community/GLM-4.7-Flash-8bit
  • Q8_0
    • Ollamaollama run hf.co/unsloth/GLM-4.7-Flash-GGUF:Q8_0
    • LM Studiolms get unsloth/GLM-4.7-Flash-GGUF@q8_0
    • llama.cppllama-server -hf unsloth/GLM-4.7-Flash-GGUF:Q8_0